Commits

Show all
Author Commit Message Labels Comments Date
Ilya Osadchiy
numpy: something on multidimensions
Branches
numpy-multidim-exp
Antonio Cuni
kill pdb and fish the config from the translator
Antonio Cuni
merge heads
Antonio Cuni
don't add a dependency on libffi unless we enable _ffi
Antonio Cuni
kill the jit_ffi attributes, which is a leftover of the now outdated 92141efc20c2
Alex Gaynor
Added signature tests for slicing.
Armin Rigo
Fix for some Linuxes on which there is really no libffi.a, but just libffi.so.
Armin Rigo
(mvt) Fix for OS/X.
Maciej Fijalkowski
add tid also to SizeDescr to silence rtyper warning
Maciej Fijalkowski
merge default
Ilya Osadchiy
merge
Ilya Osadchiy
numpy: addit test_zjit for slicing
Ilya Osadchiy
merge upstream
Ilya Osadchiy
numpy: some tests for slices
Ilya Osadchiy
numpy: signature per SingleDimSlice class
Ilya Osadchiy
merge upstream
Ilya Osadchiy
numpy: deleting sources when computation ends in VirtualArray
Ilya Osadchiy
numpy: receiving slice object as argument of descr_getitem
Ilya Osadchiy
numpy: fixing typo in slicing
Ilya Osadchiy
micronumpy: start implementation of slicing
Benjamin Peterson
make package work on osx
Amaury Forgeot d'Arc
Fix me: trackgcroot.py fails in pypy_g_ll_math_ll_math_frexp, when compiled by msvc. This assembler snippet is copied from a run of test_asmgcroot.py, but the win32 buildbot probably stopped on a similar one.
Amaury Forgeot d'Arc
(theller) Add Windows console I/O routines to the msvcrt module
Armin Rigo
Try to fix rounding in datetime.fromtimestamp(). Mostly done by trying out reasonable-looking alternatives until found one that matches CPython's behavior. I *think* that the new test covers all cases, so it should work.
Armin Rigo
Fix, indirectly for translator.c.test.test_lladdresses
Armin Rigo
Ah sorry, re 69cadfd7c8e1. Found how to reproduce (translate -Ojit, of course). Add a "correct" fix.
Armin Rigo
Revert 69cadfd7c8e1. That looks wrong. If targetnumpy shows an annotator issue, we should understand it and fix it correctly. Unfortunately targetnumpy works for me...
Armin Rigo
I *think* it's a good idea to do this in that order instead. It matches the change of order done in the x86 jit backend, too.
Benjamin Peterson
account for new arg
Benjamin Peterson
invent_fail_descr now takes another argument
  1. Prev
  2. Next